Search

Deep convolutional neural networks(DCNN) has recently significantly pushed the state of the art of semantic image segmentation. However, as the network depth increases, the progressively lower resolution leads to a loss of feature information and a large number of segmentation errors near the image edges. This is an imperative problem for applications that require accurate segmentation such as intrusion detection of railway. To address this problem, we propose superpixel pooling DCNN combined with random walks networks (SDR-Net), a semantic segmentation model that combines improved DeepLabV3+ and random walks algorithm. Firstly, the improved DeepLabV3+ with superpixel-pooling ASPP (SPASPP) is used as the basic structure of SDR-Net to introduce image prior information and improve the image feature extraction performance. Then, all pixels of the initial segmentation are classified by a binary classification network based on D2S transformation into two categories: correctly labeled pixels and incorrectly labeled pixels. Thirdly, the correctly labeled pixels are used as seed points, and then random walk algorithm is applied to reclassify the mislabeled pixels to achieve higher precision semantic image segmentation. Experiments on PASCAL VOC2012 datasets show the effectiveness of the algorithm in this paper, and it was successfully applied to rail intrusion detection.

The rapid development of industry intelligence has promoted the importance of sorting robots, but the shortcomings of sorting tasks are also gradually obvious. The target detection of the sorting robot in the process of sorting and grasping is affected by the anchor frame setting, and its grasping efficiency and recognition effect are low. Therefore, a detection network based on an anchor frame-free regression algorithm is proposed in this study, and a positioning and grasping method of the sorting robot based on the Deep Belief Network (DBN) algorithm is proposed. The algorithm experiment shows that the prediction accuracy of the proposed anchor-free detection network in the PASCAL VOC dataset reaches 92.91, significantly higher than that of other detection networks. In addition, the sorting robot was researched and designed with high accuracy in occlusion target recognition. When the occlusion area reaches 70%, the accuracy rate is as high as 81.26%, which is far higher than other detection schemes. The above results show that in improving the sorting robot’s target recognition ability, the anchor-free detection network can significantly improve its recognition accuracy. At the same time, introducing the DBN network can improve the robot’s recognition ability for occluded targets, which is of great significance to the development of the sorting robot and the intelligent industry. [ABSTRACT FROM AUTHOR]

Semitransparent solar cells are promising power-generating windows for building integrated photovoltaics, and the color-fidelity is crucially important for their potential application. Herein, to obtain desirable neutral-colored perception while surmounting the single-junction efficiency barrier, a systematic method is introduced to design semitransparent tandem solar cells (ST-TSCs), and CsPbI2Br/PTB7-Th:IEICO-4F-based perovskite/organic tandem solar cells (P/O-TSCs) with outstanding color-fidelity are first reported. Above all, a numerical method, which includes the fluctuation factor of external quantum efficiency (EQEff) within the visible range to select the front-cell and the absorption selectivity factor within the near-infrared spectrum (SNIR) to assess the rear-cell, is established to construct the device structure. Second, the theoretical transmittance fluctuation factor within the visible range (TFFVIS) and absorption selectivity factor (S) are developed to evaluate ST-TSCs' color-fidelity and photovoltaic performance. Finally, a PTB7-Th layer is introduced to passivate the halogen vacancy defects of CsPbI2Br. As a result, the neutral-colored semitransparent P/O-TSCs with a color coordinate of (0.344, 0.349) are achieved, which possess an efficiency of 13.56%, an average visible transmittance of 26.28% and a color rendering index of 96.4. The synergistic effects of the transmittance fluctuation factor and light-absorption selectivity open the path to design and fabricate semitransparent P/O-TSCs with balanced efficiency and good color neutrality. [ABSTRACT FROM AUTHOR]